Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.programming.threads > #2364
| From | aminer <aminer@toto.net> |
|---|---|
| Newsgroups | comp.programming.threads, comp.programming |
| Subject | Re: So follow with me please... |
| Date | 2014-05-18 20:12 -0700 |
| Organization | albasani.net |
| Message-ID | <llbi9e$2b6$1@news.albasani.net> (permalink) |
| References | <llbh2a$muk$1@news.albasani.net> |
Cross-posted to 2 groups.
On 5/18/2014 7:52 PM, aminer wrote:> So my new algorithm has more parallelism than the two locks algorithm, > because many threads can execute the if "fcount1^[lastHead and > fMask].flag1=0" in parallel and hence it will higher the throughput, > because in new CPU architechtures we can have many bidirectional > cache-to-cache communications running in parallel... I mean with more and more cores and on new architectures many cache-lines of the "fcount1^[lastHead and fMask].flag1" of my new algorithm will be transfered in parallel and this will make the P/N part part of the Amdahl law smaller (P: is the parallel part, and N: is the number of cores) , so this will higher the throughput, because in new CPU architechtures we can have many bidirectional cache-to-cache communications running in parallel as i have explained it to you. Thank you, Amine Moulay Ramdane.
Back to comp.programming.threads | Previous | Next — Previous in thread | Find similar | Unroll thread
So follow with me please... aminer <aminer@toto.net> - 2014-05-18 19:51 -0700 Re: So follow with me please... aminer <aminer@toto.net> - 2014-05-18 20:12 -0700
csiph-web